Researchers recently trained a deep learning (DL) algorithm to quantify glaucomatous neuroretinal damage on fundus photographs using the minimum rim width relative to the Bruch’s membrane opening (BMO-MRW) from spectral-domain optical coherence tomography (SD-OCT) as reference. They note that the algorithm showed high accuracy for glaucoma detection and may potentially eliminate the need for human grading of optic disc photos.

This cross-sectional study evaluated 9,282 pairs of optic disc photographs and SD-OCT optic nerve head scans from 927 eyes of 490 subjects who were randomly divided into the validation plus training set (80%) and the test set (20%). A DL network was trained to predict the SD-OCT BMO-MRW global and sector values when evaluating optic disc photographs. The predictions of the DL network were then compared with the actual SD-OCT measurements. Statistical analsis was used to evaluate the ability of the network to discriminate glaucomatous from normal eyes.

The team found that the DL predictions of global BMO-MRW from all optic disc photos in the test set were highly correlated with the observed values from SD-OCT, with a mean absolute error of predictions of 27.8μm. They add that the AUCs for discriminating glaucomatous from healthy eyes with the DL predictions and actual SD-OCT global BMO-MRW measurements were 0.945 and 0.933, respectively.